Optimising mining refrigeration systems through artificial intelligence

Abstract

The deep-level mining industry needs to overcome large operating costs to remain profitable. One of the largest contributors to operating costs is a mine’s refrigeration and ventilation system, which consumes up to 28% of a mine’s total energy cost. As a result, reducing the operating cost of the refrigeration system will have a large impact on a mine’s profitability. Refrigeration optimisation is often done by applying static optimisation techniques. Although some optimisations use simulation to reduce operating costs, optimisations are not robust enough when subject to varying inlet temperatures, flows and demands. Other optimisations use efficient plant selection, but do not consider varying tariff periods, such as peak, standard and off-peak periods. This study attempts to address these shortcomings by considering varying tariff structures and using artificial intelligence (AI) to determine operating conditions. The individual refrigeration plants were characterised using historical data to obtain realistic models for each plant as well as the entire system. AI optimisation techniques were used as a solution model and applied to the characteristic model to determine the optimal operating conditions. The use of AI allowed the solution model to be implemented in dynamically automated control refrigeration systems. Manual implementation was done to validate the optimisation model as the infrastructure did not allow for automated implementation. The operation did not use the storage facilities in the system and most savings were the result of water storage utilisation during peak periods. Efficient plant selection further increased the savings, which resulted in an annual saving of R9.5 million due to the reduced energy consumption during peak tariff periods. The optimisation model requires little maintenance. The system has been operational for more than two months with no need to maintain the models. Only models and varying system limitations need to be updated according to plant maintenance or downtime. Furthermore, the savings can be maintained by implementing an automated system. The automated system will prevent a loss in savings due to human error or lack of manual implementation by operators.